State of Texas v. Biden

Before BARKSDALE, ENGELHARDT, and OLDHAM, Circuit Judges. ANDREW S. OLDHAM, Circuit Judge: This case concerns the Migrant Protection Protocols (“MPP” or the “Protocols”), which the Secretary of the Department of Homeland Security (“DHS”) created on December 20, 2018. On January 20, 2021, DHS suspended the MPP program (the “Suspension Decision”). On June 1, 2021, DHS permanently terminated MPP (the “Termination Decision”). DHS explained these two decisions in a series of increasingly lengthy memoranda; the first contained just a few sentences, while the last spanned 39 single- spaced pages. Texas and Missouri (the “States”) challenged both the Suspension Decision and the Termination Decision in federal court. After a full bench trial, the district court determined that the Termination Decision violated both the Administrative Procedure Act (the “APA”) and an immigration statute, 8 U.S.C. § 1225. The district court therefore vacated the Termination Decision and ordered DHS to implement the Protocols in good faith or to take a new agency action that complied with the law. DHS chose not to take a new agency action. It instead chose to notice an appeal and defend its Termination Decision in our court. DHS also asked us to stay the district court’s injunction while the appeal was pending. We denied that motion, and the Supreme Court affirmed our denial. The Government thereafter vigorously defended the Termination Decision before our court. Then, on the Friday before oral argument—October 29, 2021—DHS issued two more memoranda (the “October 29 Memoranda” or “Memoranda”) to explain the Termination Decision. These much longer documents purported to “re-terminate” MPP—or at the very least, promised to do so after the lifting of the district court’s injunction. A few hours later, the Government informed our court that, in its view, the October 29 Memoranda had mooted this case. Never mind that a case is moot only when the controversy between the parties is dead and gone, and the controversy between these parties is very much not dead and not gone. Never mind that the new memoranda simply reaffirmed the Termination Decision that the States had been challenging all along. And never mind that the Government’s theory of mootness would allow an administrative agency to permanently avoid judicial review by issuing an endless litany of new memos to “moot” every adverse judicial ruling. The Government boldly proclaimed that DHS’s unilateral decision to issue new memoranda required us to give DHS the same relief it had previously hoped to win on appeal—namely, vacatur of the district court’s injunction and termination of MPP. DHS’s proposed approach is as unlawful as it is illogical. Under Supreme Court and Fifth Circuit precedent, this case is nowhere near moot. And in any event, the vacatur DHS requests is an equitable remedy, which is unavailable to parties with unclean hands. The Government’s litigation tactics disqualify it from such equitable relief. The Government also raises a slew of reviewability arguments, contending that no court may ever review the Termination Decision. DHS claims the power to implement a massive policy reversal—affecting billions of dollars and countless people—simply by typing out a new Word document and posting it on the internet. No input from Congress, no ordinary rulemaking procedures, and no judicial review. We address and reject each of the Government’s reviewability arguments and determine that DHS has come nowhere close to shouldering its heavy burden to show that it can make law in a vacuum. On the merits, the Termination Decision was arbitrary and capricious under the APA. That Act, among other things, requires courts to set aside agency actions that overlook relevant issues or inadequately explain their conclusions. We anchor our analysis to a recent Supreme Court decision that applied this doctrine in the immigration context. Under that precedent, this is not a close case. The Termination Decision is independently unlawful because it violates 8 U.S.C. § 1225. That statute (among other things) requires DHS to detain aliens, pending removal proceedings, who unlawfully enter the United States and seek permission to stay. It’s true that DHS lacks the capacity to detain all such aliens. Congress, however, created a statutory safety valve to address that problem. Another part of § 1225 allows DHS to return aliens to contiguous territories, like Mexico, while removal proceedings are pending. That safety valve was the statutory basis for the Protocols. DHS’s Termination Decision was a refusal to use the statute’s safety valve. That refusal, combined with DHS’s lack of detention capacity, means DHS is not detaining the aliens that Congress required it to detain. The Government insists that a third provision (in § 1182) lets DHS parole aliens into the United States on a case-by-case basis. The idea seems to be that DHS can simply parole every alien it lacks the capacity to detain. But that solves nothing: The statute allows only case-by-case parole. Deciding to parole aliens en masse is the opposite of case-by-case decisionmaking. * * * This opinion has five parts. Part I.A, infra pages 6–10, addresses this case’s factual background. Part I.B, infra pages 10–13, summarizes its statutory background. Part II addresses our jurisdiction. We start with final agency action. Part II.A, infra pages 13–29, pinpoints the final agency action under review. The final agency action is DHS’s June 1 Termination Decision. We have jurisdiction to review that Termination Decision, rather than one or the other of DHS’s ever-growing collection of MPP memos. Then we turn to mootness in Part II.B, infra pages 29–46. The October 29 Memoranda have no present legal effect, so they can’t moot the case. See Part II.B.1, infra pages 30–32. Independently, the Government has not shown they do anything to cure the Termination Decision’s unlawfulness, so again, they can’t moot the case. See Part II.B.2, infra pages 32–39. And they constitute (at most) voluntary cessation, so yet again, they can’t moot the case. See Part II.B.3, infra pages 39–45. And ordinary appellate principles bar our review of the merits of the October 29 Memoranda in any event. See Part II.B.4, infra pages 45–46. Part II.C, infra pages 46–63, addresses the States’ standing. The district court based its standing analysis on factual findings that were not clearly erroneous. See Part II.C.1, infra pages 46–52. Given those findings and the States’ entitlement to special solicitude in the analysis, we hold the States have standing. See Part II.C.2, infra pages 52–63. Part III then addresses and rejects a host of non-jurisdictional objections to the reviewability of the Termination Decision. Part III.A, infra pages 63–65, holds the States have a cause of action. Part III.B, infra pages 65–88, holds the APA does not preclude our review of the Termination Decision. Part III.B.1, infra pages 66–68, holds the immigration statutes don’t insulate the Termination Decision from review. Part III.B.2, infra pages 68–88, holds that Heckler v. Chaney, 470 U.S. 821 (1985), does not bar review either. That’s largely because Heckler, far from forbidding judicial review of agency rules, powerfully supports it. Background principles of English and American law, the Supreme Court’s precedents, and our own court’s precedents all point toward that same conclusion. See Part III.B.2.a, infra pages 69–85. Even if Heckler applied to some rules, it wouldn’t apply to the Termination Decision. See Part III.B.2.b, infra pages 85–87. And even if Heckler were presumed to apply, that presumption would be rebutted by the clear statutory text at play in this case. See Part III.B.2.c, infra pages 87–88. Part IV, infra pages 88–106, evaluates the merits. The Termination Decision was arbitrary and capricious under the APA for all sorts of reasons, and the Government’s arguments to the contrary are meritless. See Part IV.A, infra pages 88–97. And the Decision was contrary to 8 U.S.C. § 1225. See Part IV.B, infra pages 98–106. Part V, infra pages 106–17, considers the remedy. Because the case is not moot, we will deny the Government’s motion to vacate the district court’s judgment. Even if the case were moot, the Government’s litigation tactics would require the same result. See Part V.A, infra pages 106–09. And the district court didn’t abuse its discretion by vacating the Termination Decision. See Part V.B, infra pages 109–11. Nor did it abuse its discretion by imposing a permanent injunction. See Part V.C, infra pages 111–17. In sum, we hold that the Termination Decision violates both the Administrative Procedure Act and the immigration statutes. The Government’s motion to vacate the judgment and remand for further proceedings is DENIED. The district court’s judgment is AFFIRMED. I. A. This story began on December 20, 2018. On that day, DHS implemented the MPP program in response to a surge of unlawful entries along the Nation’s southern border. See Texas v. Biden (Biden I), F. Supp. 3d , 2021 WL 3603341, at *4 (N.D. Tex. Aug. 13, 2021). Before MPP, resource constraints forced DHS to release thousands of undocumented aliens into the United States and to trust that those aliens would voluntarily appear for their removal proceedings. Under MPP, DHS instead returned certain undocumented aliens to Mexico for the duration of their removal proceedings. MPP’s goal was “to ensure that certain aliens attempting to enter the U.S. illegally or without documentation . . . will no longer be released into the country, where they often fail to file an asylum application and/or disappear before an immigration judge can determine the merits of any claim.” Id. at *5 (quotation omitted). Congress expressly authorized the MPP program by statute. See 8 U.S.C. § 1225(b)(2)(C). In December 2018, Mexico agreed to admit MPP enrollees so such aliens could be held outside the United States pending their removal proceedings. Biden I, 2021 WL 3603341, at *5. In January 2019, “DHS began implementing MPP, initially in San Diego, California, then El Paso, Texas, and Calexico, California, and then nationwide.” Ibid. In February 2019, U.S. Immigration and Customs Enforcement issued guidance on MPP to its field offices, anticipating the expansion of MPP across the border. Ibid. “By December 31, 2020, DHS had enrolled 68,039 aliens in . . . MPP.” Ibid. On January 8, 2021, DHS and Texas finalized a Memorandum of Understanding (the “Agreement”). Id. at *6–7. The Agreement required Texas to provide information and assist DHS to “perform its border security, legal immigration, immigration enforcement, and national security missions.” Id. at *6 (quotation omitted). In return, DHS agreed to consult Texas and consider its views before taking actions that could modify immigration enforcement. See id. at *6–7. DHS also agreed to “provide Texas with 180 days’ written notice of any proposed action subject to the consultation requirement,” id. at *7 (quotation omitted), so that Texas would have an opportunity to comment on the proposal. The Agreement further required DHS to consider Texas’s input “in good faith” and, if it decided to reject Texas’s input, to “provide a detailed written explanation” of its reasons for doing so. Ibid. (emphasis omitted). On Inauguration Day, the Biden Administration announced its Suspension Decision. In it, DHS stated that it would suspend further enrollments in MPP. DHS’s Acting Secretary wrote, “[e]ffective January 21, 2021, the Department will suspend new enrollments in the Migrant Protection Protocols (MPP), pending further review of the program. Aliens who are not already enrolled in MPP should be processed under other existing legal authorities.” Ibid. (quotation omitted). On February 2, 2021, DHS sent a letter to Texas purporting to terminate the Agreement “effective immediately.” Ibid. Because it believed that the letter did not comply with the Agreement’s required consultation- and-explanation procedures, Texas interpreted the February 2 letter “as a notice of intent to terminate” the Agreement. Ibid. On April 13, 2021, the States sued, challenging DHS’s Suspension Decision. Id. at *1. The States claimed that the Suspension Decision violated the APA, the Immigration and Nationality Act (“INA”), the Constitution, and the Agreement. See ibid. On May 14, the States moved for a preliminary injunction that would enjoin DHS from enforcing and implementing the Suspension Decision. Ibid. On June 1, 2021, before briefing on the preliminary injunction had concluded, DHS announced its Termination Decision. The district court concluded that the Termination Decision mooted the States’ complaint about the Suspension Decision, and the court allowed the States to amend their complaint and to file a new preliminary injunction motion. Ibid. The parties agreed to consolidate the preliminary injunction hearing with the trial on the merits under Federal Rule of Civil Procedure 65(a)(2). Id. at *2. Following the bench trial, the district court issued a 53-page memorandum opinion and order, concluding that the States were entitled to relief on their APA and statutory claims. Biden I, 2021 WL 3603341. The court made many findings of fact that will be relevant here. See Part II.C.1, infra pages 46–52. Based on those findings, the court concluded that the States had Article III standing, that the Termination Decision constituted final and reviewable agency action under the APA, and that the States were within the INA’s zone of interests. Biden I, 2021 WL 3603341, at *11, 13, 17. The court then concluded that DHS’s Termination Decision was arbitrary and capricious, and therefore unlawful, under the APA. Id. at *17–18. It also concluded terminating MPP, in circumstances where DHS lacked adequate detention capacity, caused DHS to violate 8 U.S.C. § 1225(b). Id. at *22–23. Based on those conclusions, the district court vacated the Termination Decision, “permanently enjoined and restrained [DHS] from implementing or enforcing” it, and ordered DHS “to enforce and implement MPP in good faith until such a time as it has been lawfully rescinded in compliance with the APA and until such a time as the federal government has sufficient detention capacity to detain all aliens subject to mandatory detention under Section [1225] without releasing any aliens because of a lack of detention resources.” Id. at *27 (emphases omitted). DHS appealed. On August 17, 2021, the Government requested an emergency stay. See FED. R. APP. P. 8. A panel of our court denied that request and expedited the appeal. Texas v. Biden (Biden II), 10 F.4th 538, 560– 61 (5th Cir. 2021) (per curiam). The Supreme Court affirmed that denial. Biden v. Texas (Biden III), S. Ct. , 2021 WL 3732667 (Aug. 24, 2021) (mem.). On September 29, DHS announced its intention “to issue [a] new memo terminating MPP.” DEP’T OF HOMELAND SEC., DHS ANNOUNCES INTENTION TO ISSUE NEW MEMO TERMINATING MPP (2021), https://perma.cc/MM95-6KUD, screenshotted at infra page 25. On October 29, on the Friday before our court was set to hear oral argument, DHS issued two new memoranda (collectively, the “October 29 Memoranda” or “Memoranda”). See DEP’T OF HOMELAND SEC., TERMINATION OF THE MIGRANT PROTECTION PROTOCOLS (2021) (“October 29 Cover Memorandum”), https://perma.cc/45CS-DRHR; DEP’T OF HOMELAND SEC., EXPLANATION OF THE DECISION TO TERMINATE THE MIGRANT PROTECTION PROTOCOLS (2021) (“October 29 Explanation Memorandum”), https://perma.cc/4KT6- T82Z. The October 29 Memoranda did not purport to alter the Termination Decision in any way; they merely offered additional reasons for it. Hours after the release of the October 29 Memoranda, the Government filed a 26-page Suggestion of Mootness and Opposed Motion to Vacate the Judgment Below and Remand for Further Proceedings (“Suggestion of Mootness”). It argued the October 29 Memoranda mooted this appeal, and it moved our court to vacate the district court’s judgment (and injunction) and remand for further proceedings. See United States v. Munsingwear, Inc., 340 U.S. 36, 39–40 (1950) (explaining the propriety of this remedy for certain cases mooted on appeal). In the alternative, the Government asked us to hold this appeal in abeyance with respect to the § 1225 claim and remand the APA portion of the appeal to the district court, with instructions to vacate and reconsider that part of the opinion. We carried those motions with the case and gave each party additional time at oral argument to address the issue. We deny the Government’s motions in Part II.B, infra pages 29–46, and Part V.A, infra pages 106–109. B. The two statutory provisions at the heart of this case come from 8 U.S.C. § 1225(b)(2). Section 1225(b)(2)(A) provides: Subject to subparagraphs (B) and (C), in the case of an alien who is an applicant for admission, if the examining immigration officer determines that an alien seeking admission is not clearly and beyond a doubt entitled to be admitted, the alien shall be detained for a proceeding under section 1229a of this title. Section 1225(b)(2)(C) provides: In the case of an alien described in subparagraph (A) who is arriving on land (whether or not at a designated port of arrival) from a foreign territory contiguous to the United States, the Attorney General may return the alien to that territory pending a proceeding under section 1229a of this title. These provisions apply, by their terms, to “applicant[s] for admission”— that is, to aliens who are seeking entry into the United States. The former provides the default rule: Aliens who are “not clearly and beyond a doubt entitled to be admitted . . . shall be detained” while removal proceedings are pending. § 1225(b)(2)(A); see also Jennings v. Rodriguez, 138 S. Ct. 830, 842 (2018) (“Read most naturally, §§ 1225(b)(1) and (b)(2) thus mandate detention of applicants for admission until certain proceedings have concluded.”). And the latter explains one permissible alternative to detention—return to a contiguous foreign territory. § 1225(b)(2)(C). Parole is also relevant to this case. Section 1182(d)(5) both grants discretion to parole certain aliens and limits that discretion in important ways. See Jennings, 138 S. Ct. at 837 (explaining the connection between this provision and § 1225(b) detention). Parole began as an administrative invention that allowed aliens in certain circumstances to remain on U.S. soil without formal admission. See T. ALEXANDER ALEINIKOFF ET AL., IMMIGRATION AND CITIZENSHIP: PROCESS AND POLICY 299 (9th ed. 2021). Congress codified the practice when it initially enacted the Immigration and Nationality Act (the “INA”) in 1952, giving the Attorney General discretion to “parole into the United States temporarily under such conditions as he may prescribe . . . any alien applying for admission to the United States.” Immigration and Nationality Act, Pub. L. No. 82-414, 66 Stat. 163, 188 (1952). Throughout the mid-twentieth century, the executive branch on multiple occasions purported to use the parole power to bring in large groups of immigrants. See ALEINIKOFF ET AL., supra, at 300. In response, Congress twice amended 8 U.S.C. § 1182(d)(5) to limit the scope of the parole power and prevent the executive branch from using it as a programmatic policy tool. First, in the Refugee Act of 1980, Congress added § 1182(d)(5)(B), which prevents the executive branch from paroling refugees unless “compelling reasons in the public interest with respect to that particular alien require” parole. Pub. L. No. 96-212, 94 Stat. 102, 108. Second, in the Illegal Immigration Reform and Immigrant Responsibility Act of 1996 (“IIRIRA”), Congress amended § 1182(d)(5)(A) by providing that parole may be granted “only on a case-by-case basis for urgent humanitarian reasons or significant public benefit.” Pub. L. No. 104-208, 110 Stat. 3009, 3009–689 (emphasis added). As it stands today, then, the § 1182(d)(5) parole power gives the executive branch a limited authority to permit incoming aliens to stay in the United States without formal authorization when their particular cases demonstrate an urgent humanitarian need or that their presence will significantly benefit the public. The power must be exercised on a case-by- case basis. Quintessential modern uses of the parole power include, for example, paroling aliens who do not qualify for an admission category but have an urgent need for medical care in the United States and paroling aliens who qualify for a visa but are waiting for it to become available. ALEINIKOFF ET AL., supra, at 299. Parole terminates “when the purposes of . . . parole shall, in the opinion of the Attorney General, have been served.” 8 U.S.C. § 1182(d)(5)(A). At that point, DHS must treat the former parolee “in the same manner as . . . any other applicant for admission to the United States.” Ibid. The second source of parole power is in § 1226(a). Section 1226(a) provides its own detention-and-parole scheme that applies to aliens who have already entered the United States—in contradistinction to the applicants for admission covered by § 1225(b)(2) and § 1182(d)(5). See Jennings, 138 S. Ct. at 837 (explaining § 1226 “generally governs the process of arresting and detaining” inadmissible aliens who are already “inside the United States”); see also Part IV.B, infra pages 98–106 (explaining the distinction). This provision generally requires DHS to obtain an administrative warrant before arrest. See § 1226(a) (“On a warrant issued by the Attorney General, an alien may be arrested and detained pending a decision on whether the alien is to be removed from the United States.”). DHS may release such “arrested alien[s]” on either bond (at least $1,500) or conditional parole (subject to restrictions). § 1226(a)(2)–(3). II. We start, as always, with jurisdiction. First, we hold DHS’s June 1 Termination Decision constitutes “final agency action.” Second, we hold DHS’s October 29 Memoranda did not moot this case. Third, we hold the States have standing to sue. A. The APA allows judicial review of “final agency action for which there is no other adequate remedy in a court.” 5 U.S.C. § 704. For an agency action to qualify as final, the action must (1) “mark[] the consummation of the agency’s decisionmaking process” and (2) either determine “rights or obligations [or produce] legal consequences.” Texas v. EEOC, 933 F.3d 433, 441 (5th Cir. 2019) (quoting Bennett v. Spear, 520 U.S. 154, 177–78 (1997)). Our circuit considers this “a jurisdictional prerequisite of judicial review.” Louisiana v. U.S. Army Corps of Eng’rs, 834 F.3d 574, 584 (5th Cir. 2016). We begin by analyzing the June 1 Termination Decision on its own terms. We conclude the Decision was final agency action. Then, we address a new finality argument—based on the October 29 Memoranda—that the Government raises for the first time in its Suggestion of Mootness. 1. The Government says the Termination Decision didn’t consummate DHS’s decisionmaking process. That’s because a policy statement isn’t final until the agency applies it “in a particular situation” to an affected person or entity. Nat’l Mining Ass’n v. McCarthy, 758 F.3d 243, 253 (D.C. Cir. 2014) (quotation omitted). And the Government hints DHS has not yet made “the return decision” in any “individual case.” It’s hard to tell what this means. Perhaps the Government is suggesting that, somehow, DHS’s Termination Decision has not affected a single undocumented alien. But that would be absurd: DHS enrolled over 68,000 aliens in MPP when it was in effect and returned more than 55,000 of those to Mexico. Biden I, 2021 WL 3603341, at *5–6. As the district court found, MPP’s termination altered that status quo and caused DHS to return fewer aliens to Mexico (and to instead release and/or parole them into the United States). Id. at *8. If MPP’s termination did nothing at all to change the outcome in any given case, one can only imagine why the Government bothered to appeal a district court decision about an entirely nugatory policy choice. We therefore conclude that the Termination Decision was the consummation of the agency’s decisionmaking process. Likewise with the second finality prong. The Termination Decision is final agency action under the principle that, “where agency action withdraws an entity’s previously-held discretion, that action alters the legal regime, binds the entity, and thus qualifies as final agency action” under the APA. EEOC, 933 F.3d at 442 (quotation omitted). DHS withdrew its officers’ previously existing discretion on June 1 when it directed “DHS personnel, effective immediately, to take all appropriate actions to terminate MPP, including taking all steps necessary to rescind implementing guidance and other directives issued to carry out MPP.” DHS also explicitly refused to “maintain[] the status quo or [to resume] new enrollments in the program.” The Termination Decision thus bound DHS staff by forbidding them to continue the program in any way from that moment on. See id. at 441 (reiterating that binding effect upon the agency is the key inquiry and explaining that “[w]hether an action binds the agency is evident if it either appears on its face to be binding[] or is applied by the agency in a way that indicates it is binding” (quotation omitted)). The Government again responds by wishing the law said otherwise. On its view, terminating MPP can’t be final agency action because the termination “did not end DHS’s statutory authority under Section 1225(b)(2)(C) to conduct returns.” So the Government doesn’t seem to contest that the Termination Decision binds DHS staff. Instead, the idea seems to be that agency action is never final in virtue of its binding effect on agency staff—but instead is final only if the agency as a whole permanently swears off the entirety of its statutory discretion. We are aware of no case from any court that supports that sweeping proposition. And our decision in EEOC forecloses it. That case explicitly centered its finality analysis on whether “the agency’s action binds its staff.” 933 F.3d at 442 (emphasis added). Thus, our court based its holding (“that the Guidance binds EEOC”) largely on the fact that the “Guidance” in question, despite its name, bound EEOC staff. See id. at 443. The court also discussed the Guidance’s de facto creation of safe harbors for private parties. Ibid. What it did not consider is whether the EEOC could revoke its Guidance in the future. As we explained in the Heckler context in Texas v. United States (DAPA), 809 F.3d 134 (5th Cir. 2015), “[r]evocability . . . is not the touchstone for whether agency action is reviewable.” Id. at 167. And for good reason. The Government’s rule would render any agency action nonreviewable so long as the agency retained its power to undo that action or otherwise alter it in the future. That accords with neither common sense nor the law. See Sackett v. EPA, 566 U.S. 120, 127–28 (2012) (concluding the EPA’s issuance of a compliance order was final agency action and noting, “[t]he mere possibility that an agency might reconsider in light of ‘informal discussion’ and invited contentions of inaccuracy does not suffice to make an otherwise final agency action nonfinal”); cf. FCC v. Fox Television Stations, Inc., 556 U.S. 502, 513–18 (2009) (reviewing an agency action, without discussing finality, in precisely a situation where the agency had taken the opposite stance in the past). Thus, the mere fact the Termination Decision left intact DHS’s statutory authority to return aliens to contiguous territories does not undercut its finality. The Government also asserts the Termination Decision is a general policy statement—and therefore can neither determine rights nor produce obligations or legal consequences. Even if the Termination Decision is merely a “policy statement,” this argument ignores our precedent establishing that such statements can nonetheless constitute “final agency action” under the APA. See Merchs. Fast Motor Lines, Inc. v. ICC, 5 F.3d 911, 919–20 (5th Cir. 1993). The Government counters that Fast Motor Lines was a case about APA ripeness and “provided no analysis on this issue.” To the contrary, however, Fast Motor Lines reached the ripeness issue precisely because it had already concluded the agency action in question was final (despite simultaneously being a statement of policy). Id. at 920 (concluding the policy statement was final “within the meaning of 5 U.S.C. §§ 551(13) & 704” (emphasis added)). The inquiry in our circuit does not focus on labels, and it does not rely on a sharp (and false) dichotomy between statements announcing policies and final statements. The inquiry instead centers on whether the action in question determines “ rights or obligations” or creates “legal consequences.” Bennett, 520 U.S. at 178 (quotation omitted). And one way an agency can do that is by binding its own staff. That is exactly what DHS did in the Termination Decision by commanding staff to stop enrolling aliens in MPP and to terminate the program immediately. 2. In its Suggestion of Mootness, the Government now argues that the October 29 Memoranda change the picture. Even if the June 1 Termination Decision was final agency action at the time, says the Government, it lost that status when DHS issued its new Memoranda. To begin, we note that the Government could have, but did not, make this argument in its brief. The briefing obviously concluded before October 29. But the Government’s brief includes an introductory footnote that reads: “DHS has authorized us to report that the Secretary is reviewing the June 1 Memorandum and evaluating policy options regarding MPP. The result of that review could have an impact on this appeal.” So the Government knew DHS was considering a new memorandum. This would lend itself quite naturally, one would think, to an argument of the same sort the Government makes now. Yet the Government omitted the argument from its brief and instead raises it for the first time in its Suggestion of Mootness. That gives us pause. See U.S. Bancorp Mortg. Co. v. Bonner Mall P’ship, 513 U.S. 18, 27 (1994) (“To allow a party . . . to employ the secondary remedy of vacatur as a refined form of collateral attack on the judgment would—quite apart from any considerations of fairness to the parties—disturb the orderly operation of the federal judicial system.”).[1] Even so, we will consider the argument. That’s partly because the finality of agency action is a jurisdictional issue. And it’s partly because the October 29 Memoranda, which were merely possible at the briefing stage, now actually exist. First, we explain that the Government misunderstands the States’ challenge. The States are challenging DHS’s Termination Decision—not any particular memo that DHS might have written in the past or might write in the future. Second, we hold that DHS’s October 29 Memoranda did not reopen the actionable Termination Decision and are therefore not themselves final agency action. Third, we hold subsequent events can’t render a final agency action retroactively nonfinal. a. Begin with the Government’s framing of the issue. The Government treats the June 1 Memorandum as the challenged action. It then assumes that the October 29 Memoranda are a final agency action of their own. Thus, it says, the new Memoranda “demonstrate that the June 1 memorandum no longer represents the consummation of the agency’s decisionmaking process.” So even if the June 1 Memorandum was final at the moment of issuance, the October 29 Memoranda have since supplanted it as DHS’s final action under 5 U.S.C. § 704. For one thing, that framing misunderstands the nature of the challenged action.[2] The States are challenging the Termination Decision—not the June 1 Memorandum, the October 29 Memoranda, or any other memo. DHS’s Termination Decision is analogous to the judgment of a court, and its memos are analogous to a court’s opinion explicating its judgment. A judgment, not the opinion announcing that judgment, has a binding effect that settles the dispute before the court. See William Baude, The Judgment Power, 96 GEO. L.J. 1807, 1844 (2008) (describing the “historical answer” to this question: “Judgments become binding law, not opinions. Opinions merely explain the grounds for judgments, helping other people to plan and order their affairs.”). In the same way, DHS’s June 1 decision to terminate MPP had legal effect. The June 1 Memorandum—just like the October 29 Memoranda and any other subsequent memos—simply explained DHS’s decision. Thus, as common sense would indicate, the Termination Decision itself (not a memo) consummated the agency’s decisionmaking process by permanently terminating MPP. See Bennett, 520 U.S. at 177–78. The Termination Decision (not a memo) created legal consequences by stripping preexisting discretion from DHS’s own staff. See ibid.; EEOC, 933 F.3d at 443. And so the Termination Decision (not a memo) is the “final agency action” reviewable in court. 5 U.S.C. § 704. b. The October 29 Memoranda did not constitute a new and separately reviewable “final agency action.” Our holding to that effect is dictated by the well-established reopening doctrine. The D.C. Circuit developed the reopening doctrine as a way to pinpoint an agency’s final action in cases where the agency has addressed the same issue multiple times. Suppose, for example, “an agency conducts a rulemaking or adopts a policy on an issue at one time, and then in a later rulemaking restates the policy or otherwise addresses the issue again without altering the original decision.” Nat’l Ass’n of Reversionary Prop. Owners v. Surface Transp. Bd., 158 F.3d 135, 141 (D.C. Cir. 1998). What happens if the petitioner’s challenge to the agency’s action would be untimely if measured from the first agency action but timely if measured from the second?[3] The reopening doctrine provides the answer. If “the agency opened the issue up anew, and then reexamined and reaffirmed its prior decision,” the agency’s second action (the reaffirmance) is reviewable. NRDC v. EPA, 571 F.3d 1245, 1265 (D.C. Cir. 2009) (per curiam) (quotation omitted); see also Wash. All. of Tech. Workers v. DHS, 892 F.3d 332, 342 (D.C. Cir. 2018) (tying reopening to final agency action); Impro Prods., Inc. v. Block, 722 F.2d 845, 850–51 (D.C. Cir. 1983) (similar). In that event, the reaffirmance, rather than the original decision, starts the limitation period. See NRDC, 571 F.3d at 1265; Impro, 722 F.2d at 850–51. But if the agency merely reaffirmed its decision without really opening the decision back up and reconsidering it, the agency’s initial action is the only final agency action to review—so the limitation period runs from the first decision by the agency. See, e.g., Growth Energy v. EPA, 5 F.4th 1, 21–22 (D.C. Cir. 2021) (per curiam). A reopening has occurred only if “the entire context demonstrates that the agency has undertaken a serious, substantive reconsideration of the existing rule.” Id. at 21 (quotation omitted). Reversionary Property Owners v. Surface Transportation Board, 158 F.3d 135 (D.C. Cir. 1998), is the seminal case. See, e.g., CTIA—The Wireless Ass’n v. FCC, 466 F.3d 105, 110 (D.C. Cir. 2006) (treating it as such). Reversionary Property Owners concerned the Interstate Commerce Commission (the “ICC”) and its successor agency, the Surface Transportation Board (the “STB”). 158 F.3d at 137–40. Rather than owning whole railroad corridors in fee simple, railroads often hold mere rights-of-way that allow them to run tracks over others’ property. Id. at 137–38. Sometimes, railroads abandon those rights-of-way. Ibid. Before they can do so, they must get agency permission and notify the public at large by filing a “Notice of Intent.” Ibid. Sometimes, abandonments cause reversionary property interests to vest in third parties. Id. at 137, 139. In 1986, after notice and comment, the ICC adopted a “rails to trails” rule that allowed some otherwise-abandonable corridors to become public trails instead. Id. at 139. Turning a right-of-way into a trail extinguishes third- party reversionary interests in it. Ibid. (explaining this is a compensable taking). Even so, the 1986 rule didn’t require anyone to notify the holders of reversionary interests directly beforehand. Id. at 138–39. Instead, it simply required railroads to publicize a generalized notice in the Federal Register. See id. at 139. The National Association of Reversionary Property Owners (“NARPO”) believed each owner of a reversionary interest should get individualized notice before an abandonment or a rails-to-trails conversion. So NARPO asked the ICC to reopen the 1986 notice-and-comment rulemaking and reconsider that issue. Id. at 139–40. The ICC did so, but it decided not to implement the change. Ibid. And in 1990, the ICC denied NARPO’s petition for reconsideration. Ibid. All sides agreed: That was a final agency action. See id. at 141. But in 1996, after the ICC had denied NARPO’s request for a new rulemaking on the individualized-notice issue, the STB took the reins from the now-defunct ICC and issued a new Notice of Proposed Rulemaking (“NPRM”) regarding abandonment procedures. See id. at 140–41. After notice and comment, the STB’s Final Rule made some changes—but it refused to implement an individualized-notice requirement. Id. at 145–46. Thus, the D.C. Circuit had to determine whether the 1996 NPRM reopened the individualized-notice issue. The court considered three factors and held the 1996 NPRM was not a reopening. First, the court asked whether the NPRM contained either “[a]n explicit invitation to comment on a previously settled matter” or at least “[a]mbiguity” about whether the individualized-notice issue was on the table. Id. at 142. The court acknowledged the NPRM had proposed three changes to abandonment- notice procedures—including one that would require railroads to directly notify NARPO (and one other group) before abandoning a right-of-way. Id. at 141–44. It also noted the NPRM’s specific invitation for comments on “improving notice to the public.” Id. at 145 (quotation omitted). And the court acknowledged the NPRM’s broader invitation for “public comments on these proposals, and on any other areas where changes might be made . . . to streamline our abandonment regulations.” Ibid. (quotation omitted). Despite all that, the court concluded the NPRM’s text unambiguously excluded the issue of individualized notice. See ibid.; see also Growth Energy, 5 F.4th at 21–22 (an NPRM inviting comments on “any aspect of [the] rulemaking . . . did not suggest that the agency was undertaking a reconsideration of the relevant matter” (quotation omitted)). Second, the court considered the “agency’s response to comments filed by parties during [the] rulemaking.” Reversionary Prop. Owners, 158 F.3d at 142. When an agency’s “discussion of its policies and rules” regarding a given topic comes “only in response to . . . unsolicited comments,” there has likely been no reopening. Id. at 143 (quotations omitted). This is especially true when the response “merely reiterate[s]” the agency’s “ longstanding policies.” Ibid. (quotation omitted) (discussing United Transp. Union-Ill. Legis. Bd. v. Surface Transp. Bd., 132 F.3d 71 (D.C. Cir. 1998)). Accordingly, the court noted that STB’s Final Rule offered “basically the same rationale” the ICC and STB had given multiple times before. Id. at 145. For that reason, the Final Rule’s response to NARPO’s comments did not “reflect a genuine reconsideration” of the individualized-notice issue. Ibid.; see also CTIA, 466 F.3d at 112 (reaching the opposite conclusion because, among other things, the final order in question offered “two new justifications” that “constituted the [agency's] first legal rationales for its action to date”). The third factor, and arguably the court’s most important, was “the entire context of the rulemaking.” Reversionary Prop. Owners, 158 F.3d at 144 (quotation omitted) (explaining the preeminence of this consideration); see also Growth Energy, 5 F.4th at 21 (“entire context” includes “all relevant proposals and reactions of the agency” (quotation omitted)). Taken as a whole, the context did not suggest the STB was genuinely reconsidering the individualized-notice issue. Instead, the context “was one of making incremental adjustments to existing regulations and updating in light of a statute that did not call the STB’s notice provisions into question.” Reversionary Prop. Owners, 158 F.3d at 145; see also Am. Rd. & Transp. Builders Ass’n v. EPA, 588 F.3d 1109, 1115 (D.C. Cir. 2009) (conducting a similarly commonsense inquiry into “the entire context of the rulemaking” and finding no reopening (quotation omitted)). The conclusion: There was no reopening, the 1996 Final Rule wasn’t a final agency action on the individualized-notice issue, and NARPO’s suit was untimely. See Reversionary Prop. Owners, 158 F.3d at 146. Under Reversionary Property Owners and the reopening doctrine, the October 29 Memoranda did not come close to reopening DHS’s Termination Decision. First, we look for “ambiguity” in the closest thing this case has to an NPRM: DHS’s September 29 announcement of an intention to issue a new memorandum. See id. at 141–45; P & V Enters. v. U.S. Army Corps of Eng’rs, 516 F.3d 1021, 1023–27 (D.C. Cir. 2008) (demonstrating the flexibility of the Reversionary Property Owners factors by adapting them to the combination of an Advance Notice of Proposed Rulemaking and a press release). Here is a screenshot for reference: This was DHS’s first public announcement since June 1 intimating an intention to issue any new document about MPP. The title leaves nothing to the imagination, and neither does the text: Rather than announcing an intention to reconsider its Termination Decision, the announcement set forth DHS’s conclusion in unmistakable terms. The Reversionary Property Owners court found no ambiguity in an NPRM that both suggested open-mindedness about issues closely related to the one at hand and contained an explicit, broadly worded request for comments from the public. See 158 F.3d at 141–45. So how could there be any ambiguity about DHS’s September 29 announcement, which did neither? See ibid. The outcome is the same even if, arguendo, we take into account the Government’s brief. The brief notified our court on September 20 “that the Secretary is reviewing the June 1 Memorandum and evaluating policy options regarding MPP.” That’s just the kind of broad language that does not suggest a reopening. See Growth Energy, 5 F.4th at 21–22. Second, if we could, we would consider the October 29 Memoranda’s response to comments. See Reversionary Prop. Owners, 158 F.3d at 142. We can’t do that because DHS never asked for comments. That alone is enough to conclude this factor weighs against a finding of reopening. See ibid. True, the new Memoranda did respond to the Biden I court’s criticisms. See October 29 Explanation Memorandum 11–29, 36–38 (responding to the district court’s reasoning). But even if we pretended those responses were addressing comments rather than a judicial opinion, the first and third factors would outweigh this one. Cf. Am. Rd. & Transp. Builders Ass’n, 588 F.3d at 1115 (agency response given “in answer to comments received pursuant to the publication of petitioner’s own call for revisions . . . is not, without much more, sufficient to trigger the reopener doctrine” (emphasis omitted) (quotation omitted)). Third, the overall context establishes beyond doubt that DHS didn’t reopen the Termination Decision. The district court remanded to DHS “for further consideration” and went on to hold that DHS must “enforce and implement MPP . . . until such a time as it has been lawfully rescinded in compliance with the APA,” among other things. Biden I, 2021 WL 3603341, at *27 (emphasis added). In light of that decretal language, DHS announced its unambiguous intention to re-terminate MPP—without a hint of an intention to put the Termination Decision back on the chopping block and rethink things. Then its October 29 Memoranda followed through. Thus, all of DHS’s “proposals and reactions” in this case, see Growth Energy, 5 F.4th at 21–22, establish that DHS never reopened its Termination Decision—it just further defended what it had previously decided, see Reversionary Prop. Owners, 158 F.3d at 145–46. Because the October 29 Memoranda merely continued, rather than reopened, the Termination Decision, they did not embody final agency action as to that Decision. See Wash. All. of Tech. Workers, 892 F.3d at 342; Impro, 722 F.2d at 850–51. So DHS’s latest memos cannot render the June 1 Termination Decision nonfinal. c. Independently, subsequent events can’t un-finalize a final agency action. An action is either final or not, and the mere fact that the agency could—or actually does—reverse course in the future does not change that fact. Were it otherwise, only irrevocable agency actions would be final. That is exactly the rule we rejected, at the Supreme Court’s behest, just above. See Part II.A.1, supra pages 14–17; Sackett, 566 U.S. at 127–28 (“The mere possibility that an agency might reconsider in light of ‘informal discussion’ and invited contentions of inaccuracy does not suffice to make an otherwise final agency action nonfinal.”); see also Wash. All. of Tech. Workers, 892 F.3d at 342 (explaining that even if the agency’s reconsideration is a final action of its own under the reopening doctrine, the agency’s original “[r]ule was unquestionably final agency action”). The Government’s contrary view would never allow a court to make a final determination that any given agency action is final. We would be stuck in eternal limbo, waiting for the agency to give some carved-in-stone sign that the action in question is here to stay for good. That would have absurd jurisdictional consequences: Because our court views finality as a prerequisite of subject-matter jurisdiction, see Louisiana, 834 F.3d at 584, any post- judgment agency action would retroactively deprive the district court of subject-matter jurisdiction. No matter how final an agency action may appear, and no matter how sure the court’s jurisdiction to review it, the slightest agency vicissitude could destroy both finality and jurisdiction at any moment. This case illustrates the absurdity of the Government’s position. As we’ve already explained, the Termination Decision was final on June 1. See Part II.A.1, supra pages 14–17. The Termination Decision remained final when the district court reviewed it and held it unlawful on August 13. See Biden I, 2021 WL 3603341. The Termination Decision remained final when we refused to stay the district court’s decision on August 19. See Biden II, 10 F.4th 538. The Termination Decision remained final when the Supreme Court likewise refused a stay. See Biden III, 2021 WL 3732667. This tripartite judicial rebuke then prompted DHS to explain the Termination Decision anew by way of the October 29 Memoranda.[4] And those October 29 Memoranda somehow retroactively unfinalized the Termination Decision, the finality of which previously gave rise to the entire case (including the October 29 Memoranda themselves). The upshot of it all, the Government says, is that we should go back in time and hold that the district court did not have jurisdiction to start this chain of events by invaliding the Termination Decision in the first place because the future retroactively unfinalized that decision. Cf. BACK TO THE FUTURE (Universal Pictures & Amblin Ent. 1985). We are aware of no case from any court that supports the Government’s theory. Today we reject it.[5] B. Our jurisdictional inquiry also requires us to consider whether this case is moot. See Lewis v. Cont’l Bank Corp., 494 U.S. 472, 477–78 (1990) (holding mootness destroys subject-matter jurisdiction). It’s not. The Government’s Suggestion of Mootness—operating on the mistaken assumption that the agency action under review is the June 1 Memorandum rather than the underlying Termination Decision—argues as follows. The States’ supposed harms were caused by the legal defects (if any) of the June 1 Memorandum. The October 29 Memoranda superseded and rescinded the June 1 Memorandum. Just as a legislature can moot a pending appeal by amending a statute in a way that cures the statute’s defect, see id. at 478–82, so too did DHS’s October 29 Memoranda cure any legal defects in the June 1 Memorandum. See United States v. Microsoft Corp., 138 S. Ct. 1186, 1187–88 (2018) (per curiam) (holding, like Lewis, that an intervening change in a statute mooted a case). So any challenge to the June 1 Memorandum must now be moot, and the appeals court has no choice but to vacate the district court’s judgment. The Government’s stance, in more colloquial terms, is this: DHS can write a memo, litigate a case to final judgment, lose, and then immediately moot the dispute by writing a new memo overnight. Never mind that Lewis and Microsoft involved statutes instead of memos: In the Government’s view, posting a new PDF document on the internet can moot a case as easily as a statute that’s undergone bicameralism and presentment. Even better, that mootness requires this court to vacate the district court’s judgment, thus giving DHS the same relief it would have received if it had won on the merits—without the inconvenience of having to actually do so. To describe the Government’s position is to demonstrate its absurdity. We nonetheless address each of the Government’s mootness arguments in turn. We first explain that the October 29 Memoranda, on their own terms, have no present legal effect. It necessarily follows that they cannot have the legal effect of mooting this case. Second, even if the October 29 Memoranda had legal effect, the Government has not shown they cure the unlawfulness of the Termination Decision. Third, even if the October 29 Memoranda did have legal effect and did cure that unlawfulness, the new memos would constitute (at very most) voluntary cessation that does not moot the dispute. Fourth and finally, our review of the October 29 Memoranda is barred by several independent appellate principles. 1. The October 29 Memoranda cannot have the legal effect of mooting this case because those memos presently have zero legal effect. Perhaps more precisely, the memos’ legal effect is one part nullity and one part impending. The Memoranda purported to do two things: (1) “immediately supersede[] and rescind[] the June 1 Memorandum,” and (2) terminate MPP, with that termination “to be implemented as soon as practicable after a final judicial decision to vacate the . . . injunction that currently requires good faith implementation and enforcement of MPP.” October 29 Cover Memorandum 4; see also October 29 Explanation Memorandum 4–5. The October 29 Memoranda’s supposed rescission of the June 1 Termination Decision was a nullity. The district court had already vacated the Termination Decision under 5 U.S.C. § 706, which empowers and commands courts to “set aside” unlawful agency actions. See Biden I, 2021 WL 3603341, at *23–24 & n.12. That statutory empowerment means that, unlike a court’s decision to hold a statute unconstitutional, the district court’s vacatur rendered the June 1 Termination Decision void. See Jonathan F. Mitchell, The Writ-of-Erasure Fallacy, 104 Va. L. Rev. 933, 1014–16 (2018) (explaining this point); see also Driftless Area Land Conservancy v. Valcq, 16 F.4th 508, 522 (7th Cir. 2021) (“Vacatur [of an agency action] retroactively undoes or expunges a past [agency] action.”). So the October 29 Memoranda may have attempted to rescind DHS’s rationale for the Termination Decision, but that attempt had no effect because there was nothing to rescind. A nullity can’t moot a case. That leaves the Memoranda’s second purported effect: the re- termination of MPP. The October 29 Memoranda expressly state that the re- termination will have no effect until after the district court’s injunction has been lifted. See October 29 Explanation Memorandum 4–5. The Government offers no explanation for how a legal effect that has yet to occur could moot this case now. True, the new memos use equivocal phrasing to describe their legal effect, and sometimes this involves present-tense language. See, e.g., October 29 Cover Memorandum 4 (“I am hereby terminating MPP.”). But the fact remains that the Memoranda don’t purport to actually do anything until the injunction ends. Just as a nullity can’t spring forth from the void to moot a case, a prophesied legal effect can’t leap backward from the future to do so.[6] The Government objects that it would be strange to fault DHS for postponing its re-termination until the future. How else, the Government asks, could it have framed the October 29 Memoranda without risking contempt of the district court’s injunction? The answer, of course, is that the Government made the bed it’s attempting to not sleep in. The Government chose to (a) appeal this case, (b) act as though it’s returning to the district court under Federal Rule of Civil Procedure 60(b) (even though the appeal means the case is not before the district court), and (c) moot the very case it appeals, not by doing what the district court ordered it to do, but by refusing to confess error—all at the same time. The Government cannot use this have-its-cake-and-eat-it-too strategy to moot the case. See Part V.A, infra pages 106–09 (discussing that strategy in more detail). 2. Let’s nonetheless assume that the October 29 Memoranda have present legal effects. Even if such effects existed (they don’t), the Government has not shown the effects would cure the unlawfulness of the Termination Decision. Nor that they would eliminate the States’ ongoing injuries from that decision. Nor that they would remove our judicial power to grant relief against DHS. That’s an independent basis for concluding the case is not moot. A case is moot if “it is impossible for a court to grant any effectual relief whatever to the prevailing party.” Knox v. Serv. Emps. Int’l Union, Local 1000, 567 U.S. 298, 307 (2012) (quotation omitted). For challenges to governmental actions, that means “a case challenging a statute, executive order, or local ordinance usually becomes moot if the challenged law has expired or been repealed.” Spell v. Edwards, 962 F.3d 175, 179 (5th Cir. 2020). In Spell, we accordingly held moot a challenge to gubernatorial COVID-19 stay-at-home orders after those orders “expired by their own terms.” Ibid. With the orders expired, there was simply “nothing for us to enjoin.” Id. at 177. Likewise, the Supreme Court held moot a challenge to New York City gun rules after the City amended those rules in a way that gave the petitioners “the precise relief [they had] requested in the prayer for relief in their complaint.” N.Y. State Rifle & Pistol Ass’n v. City of New York, 140 S. Ct. 1525, 1526 (2020) (per curiam). But when a government repeals the challenged action and replaces it with something substantially similar, the injury remains. In such a case, the court can still “grant . . . effectual relief . . . to the prevailing party,” Knox, 567 U.S. at 307 (quotation omitted), and the case is not mooted. Consider Northeastern Florida Chapter of Associated General Contractors of America v. City of Jacksonville, 508 U.S. 656 (1993). There, Jacksonville adopted a “Minority Business Enterprise Participation” ordinance that required 10% of the city’s contracting budget to be “set aside” for deals with minority-owned contractors. Id. at 658–59. Non-minority contractors brought a Fourteenth Amendment challenge. See id. at 658–60 (describing the case’s procedural history). After the Court granted certiorari, the city “repealed its . . . ordinance and replaced it with an ordinance entitled ‘African–American and Women’s Business Enterprise Participation.’” Id. at 660. That program was slightly narrower and more flexible than the original, and it allowed for set-asides above or below 10%. Id. at 660–61. The city argued it had mooted the case by repealing and replacing the original ordinance. Id. at 661. The Court saw right through the city’s gamesmanship. The Court first explained that a defendant generally may not moot a case by voluntarily ceasing the challenged conduct. Id. at 661–62. But then it explained that the case at hand was even more obvious than that—because the defendant city hadn’t really ceased anything: This is an a fortiori case. There is no mere risk that Jacksonville will repeat its allegedly wrongful conduct; it has already done so. Nor does it matter that the new ordinance differs in certain respects from the old one. [The relevant voluntary-cessation precedent] does not stand for the proposition that it is only the possibility that the selfsame statute will be enacted that prevents a case from being moot; if that were the rule, a defendant could moot a case by repealing the challenged statute and replacing it with one that differs only in some insignificant respect. The gravamen of petitioner’s complaint is that its members are disadvantaged in their efforts to obtain city contracts. The new ordinance may disadvantage them to a lesser degree than the old one, but insofar as it accords preferential treatment to black- and female-owned contractors—and, in particular, insofar as its “Sheltered Market Plan” is a “set aside” by another name—it disadvantages them in the same fundamental way. Id. at 662.[7] Our court first applied City of Jacksonville in Cooper v. McBeath, 11 F.3d 547 (5th Cir. 1994). Faced with a challenge to its three-year residency requirement for liquor licenses, Texas repealed the relevant statute and replaced it with a one-year requirement. Id. at 549–50. City of Jacksonville, we held, was a perfect fit. Texas could not moot the case simply by tweaking its challenged law. See id. at 550–51. (“[T]he new one-year residency/citizenship requirement may lessen the burden placed on the Plaintiffs, but . . . the amendments’ practical effect remains the same: Plaintiffs, as non-Texans, are treated differently.”). Likewise in Opulent Life Church v. City of Holly Springs, 697 F.3d 279 (5th Cir. 2012). There, the city’s zoning ordinance allegedly “singled out churches for unfavorable treatment.” Id. at 281–82 (quotation omitted). The day before oral argument, the city repealed the challenged provision and replaced it with one that banned churches from certain properties outright. Id. at 284–85. We applied City of Jacksonville and held the case was not moot. Id. at 285–86; see also Big Tyme Invs., LLC v. Edwards, 985 F.3d 456, 464–65 (5th Cir. 2021) (holding an Equal Protection challenge to a COVID-19 bar closure not mooted even by the adoption of more lenient restrictions because the new rules “continue[d] to differentiate between bars and restaurants” (quotation omitted)). The same principle governs here. The Government says DHS’s October 29 Memoranda mooted this whole case by rescinding the June 1 Memorandum and replacing it with a new explanation for terminating MPP. As we’ve explained, the Termination Decision is at issue here, not the June 1 Memorandum. And even aside from that, the Government’s purported line between harms-caused-by-the-June-1-Memorandum and harms-caused-by- the-October-29-Memoranda is a distinction without a difference. This kind of faux-metaphysical quibbling ignores the “gravamen” of the States’ challenge to the Termination Decision. See City of Jacksonville, 508 U.S. at 662. DHS cannot moot this case by reaffirming and perpetuating the very same injury that brought the States into court. The Government offers two lines of response. First, it relies heavily on the Supreme Court’s order in Mayorkas v. Innovation Law Lab, 141 S. Ct. 2842 (2021) (mem.). That order concerned the mirror image of this case—a challenge to the creation of MPP rather than its termination. The district court enjoined MPP, and the Ninth Circuit affirmed. Innovation Law Lab v. Wolf, 951 F.3d 1073, 1095 (9th Cir. 2020). But on June 1, of course, DHS terminated MPP. So the Court vacated the Ninth Circuit’s judgment and remanded the case “with instructions to direct the District Court to vacate as moot the . . . order granting a preliminary injunction.” Innovation Law Lab, 141 S. Ct. at 2842. That reliance is very much misplaced. DHS’s policy change in Innovation Law Lab obviously gave the plaintiffs “the precise relief [they had] requested,” leaving the injunction with no work to do. See N.Y. State Rifle & Pistol, 140 S. Ct. at 1526. So it made sense for the Supreme Court to hold the case moot. See Innovation Law Lab, 141 S. Ct. at 2842. In this case, DHS’s October 29 Memoranda did nothing less than vow faithful adherence to the June 1 Termination Decision. Unlike the plaintiffs in Innovation Law Lab, the States are left with none of the relief they requested. That leaves the injunction with just as much work to do as ever. Next, the Government focuses on each of the States’ two merits challenges to the Termination Decision (based on 8 U.S.C. § 1225 and the APA). See Biden I, 2021 WL 3603341, at *22–23 (district court’s discussion of § 1225); Part IV.B, infra pages 98–106 (our analysis of § 1225); see also Biden I, 2021 WL 3603341, at *17–22 (district court’s discussion of the APA); Part IV.A, infra pages 88–97 (our discussion of the APA). It argues the October 29 Memoranda change the situation enough to moot the case. As for § 1225, the Government points out that DHS’s new Memoranda invoke deference under Chevron, USA, Inc. v. NRDC, 467 U.S. 837 (1984), to justify paroling any and every alien DHS lacks the capacity to detain. See October 29 Explanation Memorandum 28 (citing Chevron). So Chevron deference, which wasn’t at play before, is relevant now. And because the district court’s § 1225 reasoning relied in part on the idea that paroling all above-capacity aliens would be impermissible under § 1182(d)(5)(A), see Biden I, 2021 WL 3603341, at *22 n.11, the Government argues, the § 1225 issue is now moot. A creative move, but Chevron was as available before October 29 as it is today. The Government’s own brief points out that “DHS has long interpreted Section 1182(d)(5) to authorize parole of noncitizens who present neither a security risk [n]or a risk of absconding and whose continued detention is not in the public interest.” (Emphasis added and quotation omitted.) In fact, the American Civil Liberties Union (the “ACLU”) raised Chevron deference in an August 17 amicus curiae brief filed in our court. The brief pointed to the longstanding DHS regulation in 8 C.F.R. § 212.5(b), arguing that the regulation is a broad, deference-worthy interpretation of 8 U.S.C. § 1182(d)(5)’s parole power. And if the district court had properly deferred to that interpretation, said the ACLU, it would have realized that releasing all over-capacity aliens fits within the statutory “case-by-case basis” limitation on parole. The Government thus forfeited the Chevron issue by failing to mention it in its brief. See HollyFrontier Cheyenne Refin., LLC v. Renewable Fuels Ass’n, 141 S. Ct. 2172, 2180 (2021) (“[T]he government is not invoking Chevron. We therefore decline to consider whether any deference might be due its regulation.” (quotation omitted)); Tiger Lily, LLC v. HUD, 5 F.4th 666, 669 (6th Cir. 2021) (“Notably, the government does not ask us to grant Chevron deference to its interpretation of the relevant statute. ‘We therefore decline to consider whether any deference might be due.’” (quoting HollyFrontier, 141 S. Ct. at 2180)); cf. Ortiz v. McDonough, 6 F.4th 1267, 1275– 76 (Fed. Cir. 2021) (applying the same rule to deference under Auer v. Robbins, 519 U.S. 452 (1997)). In fact, it did not even raise the issue before the district court. See generally Biden I, 2021 WL 3603341. It now seeks to use the Suggestion of Mootness as a back door to undo those omissions. It may not do so. See Bancorp, 513 U.S. at 27 (concluding that a motion for Munsingwear vacatur is not a means to “collateral[ly] attack” the judgment); accord Hous. Chron. Publ’g Co. v. City of League City, 488 F.3d 613, 619 (5th Cir. 2007); see also Part IV.B, infra pages 98–106 (analyzing the statutory issue without regard to Chevron). And because Chevron was relevant to this case, if at all, before the October 29 Memoranda, the doctrine has no bearing at all on mootness. As for the APA, the Government argues the October 29 Memoranda alleviate the States’ injuries. The idea is that, even if the June 1 Termination Decision was arbitrary and capricious, the October 29 Memoranda are not. Thus, says the Government, DHS has fixed the problem the States complain of. Again, no. The Government has not shown the October 29 Memoranda actually cure the States’ APA-based injuries. For example, the Suggestion of Mootness’s glowing description of the October 29 Memoranda offers no analysis whatsoever on whether they are post hoc rationalizations under the demanding standard announced by the Supreme Court. See DHS v. Regents of the Univ. of Cal., 140 S. Ct. 1891, 1904–05, 1907–09 (2020) (describing a multiple-memorandum agency process strikingly similar to the process here and concluding the later memorandum could “be viewed only as impermissible post hoc rationalization[]“). Nor does the Government explain how the October 29 Memoranda, which are not final agency action of their own under the reopening doctrine, can be anything more than post hoc rationalizations of the Termination Decision. If the October 29 Memoranda are post hoc rationalizations, they are powerless to cure the June 1 Termination Decision’s problems. See Part IV.A, infra pages 88–97 (explaining the Termination Decision was arbitrary and capricious under Regents); Regents, 140 S. Ct. at 1907–09 (explaining “post hoc rationalizations . . . are not properly before us”). We need not decide that issue here. We hold only that the Government has not carried its “formidable burden” of showing that the October 29 Memoranda remove the States’ injuries by curing the Termination Decision’s APA defects. See Already, LLC v. Nike, Inc., 568 U.S. 85, 90–91 (2013) (quotation omitted) (laying out the burden for a party attempting to show its injurious conduct will not recur); City of Jacksonville, 508 U.S. at 662 (explaining that a showing that the injury is no longer occurring is a prerequisite to showing injurious conduct will not recur). 3. Even if the October 29 Memoranda had legal effects, and even if those legal effects cured the unlawfulness of the Termination Decision, the new memos would constitute at most a voluntary cessation of unlawfulness. Again, that’s an independent basis for holding the case is not moot. The voluntary-cessation rule is well settled: “[A] defendant cannot automatically moot a case simply by ending its allegedly unlawful conduct once sued.” Spell, 962 F.3d at 179 (quotation omitted). Were it otherwise, the Supreme Court has explained, “a defendant could engage in unlawful conduct, stop when sued to have the case declared moot, then pick up where he left off, repeating this cycle until he achieves all his unlawful ends.” Nike, 568 U.S. at 91. And “a defendant claiming that its voluntary compliance moots a case bears the formidable burden of showing that it is absolutely clear the allegedly wrongful behavior could not reasonably be expected to recur.” Ibid. (quotation omitted). The inquiry centers on “whether the defendant’s actions are ‘litigation posturing’ or whether the controversy is actually extinguished.” Yarls v. Bunton, 905 F.3d 905, 910 (5th Cir. 2018). Our court applies this same test in a slightly modified way when the defendant is a governmental entity. In such cases, “[w]ithout evidence to the contrary, we assume that formally announced changes to official governmental policy are not mere litigation posturing.” Sossamon v. Lone Star State of Tex., 560 F.3d 316, 325 (5th Cir. 2009), aff’d on other grounds sub nom. Sossamon v. Texas, 563 U.S. 277 (2011). In Speech First, Inc. v. Fenves, we explained three factors that can overcome the presumption. 979 F.3d 319 (5th Cir. 2020); see also id. at 328–29 (assuming “arguendo” that the presumption applies to public universities and analyzing accordingly). They are: “ (1) the absence of a controlling statement of future intention [not to repeat the challenged policy]; (2) the suspicious timing of the change; and (3) the [governmental entity's] continued defense of the challenged polic[y]” after the supposedly mooting event. Id. at 328. If all three factors obtain, the case isn’t moot. See id. at 328–29 (declining to decide whether fewer than three will suffice). This case fits Fenves like a glove. DHS has repeatedly exhibited gamesmanship in its decisionmaking. DHS first announced it was suspending MPP on Inauguration Day 2021. See Biden I, 2021 WL 3603341, at *7 (noting the suspension went into effect the next day, on January 21). As the district court pointed out: Since that day, DHS has not offered a single justification for suspending new enrollments in the program during the period of [its review of MPP]. Indeed, when the original administrative record was filed [in district court] prior to the June 1 Memorandum’s issuance, it contained only a single document — the January 20 Memorandum. There was no cost- benefit analysis or any sort of reasoned decisionmaking for a court to review. Id. at *8 (citation omitted) (emphasis added). The States challenged that Suspension Decision on April 13. They “alleged that DHS’s two-sentence, three-line memorandum” violated the APA and § 1225, among other things. Id. at *1 (quotation omitted). On May 14, the States moved for a preliminary injunction against the Suspension Decision. Ibid. In the midst of briefing, DHS tried—successfully—to moot that challenge. This by way of its June 1 Termination Decision, which permanently ended MPP. See ibid. The district court held this mooted the States’ challenge to the Suspension Decision, thus allowing the Government to avoid any responsibility for its completely unreasoned, two-sentence decision that started this whole case. Ibid. After the district court allowed the States to replead a challenge to the Termination Decision, DHS threw another last-minute wrench into the bench trial. See id. at *2. At least three weeks before the trial was scheduled to begin, DHS became aware the administrative record was missing a key document: DHS’s own 2019 assessment of MPP, which judged the policy to be a success. Ibid.; id. at *5 (describing the assessment). Despite the advance notice, DHS waited until two days before the one-day bench trial to add it to the record. Id. at *2. The States claimed unfair surprise and moved to have the addition excluded, see ECF No. 80, but the district court denied that motion, see ECF No. 85. It pointed out: “Defendants even waited until 3:27pm two days before the [trial] to file the corrected Administrative Record, despite the declaration of the custodian [explaining that DHS had noticed the omission] being electronically signed at 5:14 p.m. Eastern time the day before.” Id. at 2. This behavior, said the district court, came “perilously close to undermining the presumption of administrative regularity” courts normally accord to agency procedures. Biden I, 2021 WL 3603341, at *2 (quoting ECF No. 85 at 3). In the end, despite the States’ limited opportunity to tailor their case to the inclusion of the 2019 assessment, the assessment played a significant role in the district court’s analysis. See id. at *19 (“By ignoring its own previous assessment on the importance of deterring meritless asylum applications without a reasoned analysis for the change, [DHS] acted arbitrarily and capriciously.” (quotation omitted)). DHS continued its tactics on appeal. After we denied its motion for a stay, DHS announced its intention to issue a new memorandum. See Biden II, 10 F.4th 538 (decided August 19); DHS ANNOUNCES INTENTION TO ISSUE NEW MEMO TERMINATING MPP (posted September 29), screenshotted at supra page 25. On the Friday before oral argument— October 29—DHS issued its new Memoranda. Around 4:30 p.m. that Friday, the Government filed its 26-page Suggestion of Mootness. Those facts easily satisfy all three Fenves factors. First, DHS “has not issued a controlling statement of future intention” to refrain from repeating MPP’s termination. Fenves, 979 F.3d at 328–29. In Fenves, the University president, “in his official capacity, represent[ed] in his brief that the University has no plans to, and will not, reenact the [challenged] policies.” Id. at 328 (quotation omitted). The court held that wasn’t enough: Only “sworn testimony” from someone with “control” over the relevant policy choice would suffice. Id. at 328–29 (quotation omitted). This case is even more clear-cut. The Government’s Suggestion of Mootness doesn’t even claim that DHS has forsworn further memos on this topic. And there’s certainly nothing close to “sworn testimony” establishing such a commitment. Second, “the timing of [DHS's] policy amendments is at least as suspicious as was the timing of the changes in” Fenves. Id. at 329. In Fenves, the timing was “suspicious” because the University only began reviewing its policies after it lost in district court. Ibid. And the “changes were first announced only in the University’s appellate brief.” Ibid. Here, as in Fenves, DHS started reviewing its policy only after losing in district court. And unlike Fenves, DHS made the change the Friday before oral argument—long after briefing had concluded. That timing, combined with DHS’s pattern of belated shifts and its eleventh-hour mooting of the States’ original challenge, is more than a little “suspicious.” Ibid. Third, DHS “continues to defend the original policies . . . as it did in the district court.” Ibid. DHS’s original stance, expressed in its briefing, was that the June 1 Termination Decision was entirely defensible and legal. In no way does the Suggestion of Mootness alter that stance. Nor do the October 29 Memoranda themselves. And if any doubt remained on this score, oral argument would remove it. When asked whether “the Government believe[s] that the June 1 Memo was a lawful exercise of government power,” the Government’s counsel responded: “Yes, your honor.” Oral Argument at 55:46–55:54. Even after giving DHS “some solicitude” in the voluntary-cessation analysis, Sossamon, 560 F.3d at 325, we hold this case is not moot. Each of the three Fenves factors is at least as obvious here as in Fenves itself. DHS has therefore not borne its “formidable burden of showing that it is absolutely clear the allegedly wrongful behavior could not reasonably be expected to recur.” Nike, 568 U.S. at 91 (quotation omitted). Instead of trying to shoulder that burden, the Government asserts the Memoranda can’t possibly fit into the voluntary-cessation doctrine because DHS issued them in response to the district court’s remand order. That’s incorrect because it ignores the fundamental one-court-at-a-time rule. “The general rule is that a case can exist only in one court at a time, and a notice of appeal permanently transfers the case to us until we send it back.” United States v. Lucero, 755 F. App’x 384, 386 (5th Cir. 2018) (per curiam); see also Griggs v. Provident Consumer Disc. Co., 459 U.S. 56, 58 (1982) (per curiam) (“The filing of a notice of appeal is an event of jurisdictional significance—it confers jurisdiction on the court of appeals and divests the district court of its control over those aspects of the case involved in the appeal.”). That same principle applies when an agency notices an appeal instead of accepting a remand order. Thus, if the Government wanted the October 29 Memoranda to be assessed as a response to the district court’s remand, it should have voluntarily dismissed this appeal and asked the district court for relief from the judgment. See FED. R. APP. P. 42(b) (allowing for voluntary dismissal); FED. R. CIV. P. 60(b) (providing a mechanism for a party to seek relief from a judgment); Part V.A, infra pages 106–09 (discussing DHS’s attempt to have it both ways at once in this case). That court’s disposition of such a motion, of course, would have been an appealable final decision. See, e.g., Brumfield v. La. State Bd. of Educ., 806 F.3d 289, 296 (5th Cir. 2015) (“[T]he district court’s denial of the 60(b)(4) motion amounts to a refusal to dissolve an injunction, making the denial appealable under this court’s precedent.”). Such an approach would have run parallel to DHS’s path in Regents itself, where a post-remand DHS returned to the district court with its second memorandum, waited for the district court’s ruling, and appealed that ruling. 140 S. Ct. at 1904–05 (explaining that, before appealing, “[t]he Government asked the D. C. District Court to revise its prior order in light of [the new memorandum], but the court declined”). That’s what allowed the appeals courts in the Regents litigation to proceed with the benefit of full, first-instance review from the district court on the merits of both sets of agency documents. The Government was entirely free, of course, to appeal when it did. But it may not invoke the timing of its own appeal to avoid the voluntary- cessation doctrine. Just as a litigant cannot notice an appeal and then continue litigating the case in the district court, an agency cannot notice an appeal and then act as if it had accepted the remand order. 4. Finally, and in any event, independent principles of appellate law prohibit the Government’s efforts to inject the October 29 Memoranda into this case at the eleventh hour. Three bear emphasis. First and foremost is the record rule. Immediately after empowering courts to review agency action, the APA commands: “In making the foregoing determinations, the court shall review the whole record or those parts of it cited by a party.” 5 U.S.C. § 706; SEC v. Chenery Corp., 318 U.S. 80, 87 (1943) (“The grounds upon which an administrative order must be judged are those upon which the record discloses that its action was based.”). That rule applies not only to arbitrary-and-capricious review, see § 706(2)(A), but also to review for compliance with statutes, see § 706(2)(C). Thus, we will apply the law to the facts based on the agency record as it stood on the date of the Termination Decision—June 1. Second, the States challenged the June 1 Termination Decision in district court. They did not challenge the October 29 Memoranda, which obviously did not exist at the time of the district court proceedings. This is an appeal from the district court’s disposition of the States’ challenge, and the merits of DHS’s actions on October 29 are not before us. Indeed, because the reopening doctrine establishes that the October 29 Memoranda embodied no final agency action, see Part II.A.2.b, supra pages 20–27, we do not have jurisdiction to decide those merits. Third, the general rule is that “we are a court of review, not of first view.” Cutter v. Wilkinson, 544 U.S. 709, 718 n.7 (2005); see also Landry’s, Inc. v. Ins. Co. of the State of Pa., 4 F.4th 366, 372 n.4 (5th Cir. 2021). That rule counsels against considering the merits of the October 29 Memoranda before a district court has done so. Cf. Planned Parenthood of Greater Wash. & N. Idaho v. HHS, 946 F.3d 1100, 1110–15 (9th Cir. 2020) (holding a 2019 agency funding allocation didn’t moot a challenge to the 2018 version of the same allocation, deciding the merits without regard to the 2019 allocation, and declining to address most issues not considered by the district court).[8] C. Now, standing. Several factual findings were central to the district court’s standing analysis, and they will be central to ours as well. So we begin by reviewing those findings for clear error, and we find none. Then we conclude the States have standing to bring this suit. 1. Under clear-error review, “[i]f the district court’s view of the evidence is plausible in light of the entire record, an appellate court may not reverse even if it is convinced that it would have weighed the evidence differently in the first instance.” Brnovich v. Democratic Nat’l Comm., 141 S. Ct. 2321, 2349 (2021). That same standard applies to facts that underlie jurisdictional issues like standing. See Env’t Tex. Citizen Lobby, Inc. v. ExxonMobil Corp., 968 F.3d 357, 367 (5th Cir. 2020) (“Because this case was tried, Plaintiffs needed to prove standing by a preponderance of the evidence. A factual finding that a plaintiff met that burden is reviewed for clear error.” (citation omitted)); DeJoria v. Maghreb Petroleum Expl., S.A., 935 F.3d 381, 390 (5th Cir. 2019) (“[J]urisdiction is a legal question. But the facts that underlie a jurisdictional determination are still reviewed only for clear error.”). a. The district court’s most important finding was that MPP’s termination has increased the number of aliens released on parole into the United States, including Texas and Missouri. See Biden I, 2021 WL 3603341, at *8 (“Without MPP, Defendants are forced to release and parole aliens into the United States because Defendants simply do not have the resources to detain aliens as mandated by statute.”); see also 8 U.S.C. § 1182(d)(5) (laying out parole procedures). The court rooted that finding firmly in the evidence before it. The court noted DHS’s inadequate detention capacity, citing both a record declaration and some of DHS’s own publications on the matter. Biden I, 2021 WL 3603341, at *8–9. So it’s unsurprising that on appeal, even the Government admits DHS is “detaining at or near its capacity limits.” Next, the court pointed to evidence that “the termination of MPP has contributed to the current border surge.” Biden I, 2021 WL 3603341, at *9 (citing DHS’s own previous determinations that MPP had curbed the rate of illegal entries). And it pointed out that the number of “enforcement encounters”—that is, instances where immigration officials encounter immigrants attempting to cross the southern border without documentation—had “skyrocketed” since MPP’s termination. Ibid.; see also id. at *9 n.7 (noting a sworn statement of David Shahoulian, then the Assistant Secretary for Border and Immigration Policy at DHS, who predicted that “total [border] encounters this fiscal year [2021] are likely to be the highest ever recorded” (emphasis omitted)). Those pieces of record evidence make it eminently “plausible” that DHS’s termination of MPP has increased the total number of aliens paroled into the United States. Brnovich, 141 S. Ct. at 2349. The Government contests this fact in several ways—none of which persuades us the district court committed clear error. Broadly, the Government insists that “[t]he court cited no record evidence demonstrating that terminating MPP in fact led to an increase in the number of noncitizens released.” The district court’s record citations belie this claim. See Biden I, 2021 WL 3603341, at *8–9. So does the Government’s own brief. As discussed below, that brief faults the district court for giving DHS only two options: either detain aliens (8 U.S.C. § 1225(b)(2)(A)) or return them to Mexico (8 U.S.C. § 1225(b)(2)(C)). See Part IV.B, infra pages 98–106. Instead, says the Government, DHS has the third option of paroling aliens under 8 U.S.C. § 1182(d)(5)(A). Thus, for any given alien whose nondetention would otherwise violate § 1225, DHS can comply with the law (and was complying, before the district court’s injunction) simply by paroling that alien under § 1182. Put differently, the Government first denies DHS’s policy will increase the number of paroled aliens. Then it argues DHS is complying with the law precisely by paroling the aliens it lacks the capacity to detain rather than returning them to Mexico. That litigating position confirms the district court’s extensive record citations: MPP’s termination, combined with the lack of detention capacity, has increased and (without an injunction) will increase the total number of parolees. See Biden I, 2021 WL 3603341, at *9 (“Even if the termination of MPP played no role in the increasing number of migrants, the lack of MPP as a tool to manage the influx means that more aliens will be released and paroled into the United States as the surge continues to overwhelm DHS’s detainment capacity.”). The Government nonetheless contests the district court’s statistics as to capacity limits and offers its own statistics in their place. This is effectively a request that we re-weigh the evidence that was before the district court, and we will not do that. See Brnovich, 141 S. Ct. at 2349. The task of evaluating competing statistics is precisely the kind of task a district court is best situated to undertake. Cf. Woodfox v. Cain, 772 F.3d 358, 380 (5th Cir. 2014) (“Again, given the fact-intensive nature of the statistical inquiry, we can find no clear error in the district court’s opting to use the one-tailed and two- tailed tests.”). Third, the Government faults the district court for considering the number of encounters between immigration officials and would-be entrants at the border (called “border encounters”). The Government points out that officials might be arresting the same would-be entrants multiple times. And that could artificially inflate the number of encounters, even while the rate of illegal entries itself remains constant. So, the argument goes, the district court clearly erred by citing border encounters to conclude MPP’s termination has contributed to the border surge. This misses the mark entirely. The district court’s point was just that MPP’s termination has caused an increase in attempted illegal crossings. And the court quite reasonably used the rate of border encounters as a proxy for that rate. If illegal entry attempts increase, it’s irrelevant how many times a given entrant has tried in the past. As in any other context, a repeat offender is an offender just the same. And in all events, the Government’s false-positive theory makes sense only if the incidence of repeat entrants has increased since MPP’s termination. But it offers no such evidence, and the June 1 Memorandum itself suggests DHS made the Termination Decision with the hope that doing so would decrease the rate of repeat entry. Fourth, the Government denies that DHS ever acknowledged MPP’s effectiveness. The district court supported this proposition by reference to a DHS document that said, “MPP implementation contributes to decreasing the volume of inadmissible aliens arriving in the United States on land from Mexico.” Biden I, 2021 WL 3603341, at *5, *9 (quotation omitted). The Government points out that this quote came under the header “Metric,” and says the document was therefore doing nothing more than proposing a metric for measuring MPP’s effectiveness—not touting that effectiveness. Yes, the quote comes from a sub-header labeled “Metric.” But the prior page explains that “[t]he following are the intended goals of MPP and measurements of how those goals are currently being met.” (Emphasis added.) And just after the metric is a “Data measurement” sub-header, measuring the “Number of Aliens Enrolled in MPP.” That suggests the document was doing more than just proposing future measurements—and that instead, it was actually carrying out measurements itself. So one “plausible” “view of the evidence” is that DHS was not just proposing a metric but in fact concluding MPP had already successfully reduced illegal entries. See Brnovich, 141 S. Ct. at 2349. The district court did not clearly err by interpreting the DHS document the way it did. Last and related, the Government argues MPP was an ineffective deterrent, and that its termination therefore could not have caused an increase in illegal entries. But the district court made the contrary finding after its own consideration of the record and weighing of the evidence. See Biden I, 2021 WL 3603341, at *8–9. That finding is “plausible in light of the entire record,” Brnovich, 141 S. Ct. at 2349, and we will not disturb it on appeal. And even if the Government were correct that MPP was an ineffective deterrent, the fact remains that, according to both the record and the Government’s own brief, MPP’s termination drastically increases the proportion of incoming aliens who are paroled rather than returned to Mexico. And it is precisely that increase in paroles that causes the States’ harms. b. The district court found that the increase in parolees causes the States financial harm by way of driver’s license applications. Biden I, 2021 WL 3603341, at *9–10. More specifically, the court found both that “[a]s a result of the termination of MPP, some aliens who would have otherwise been enrolled in MPP are being released or paroled into the United States and will obtain Texas driver’s licenses” at a cost to Texas, and that “[e]ach additional customer seeking a Texas driver’s license imposes a cost on Texas.” Id. at *9. Neither finding was clearly erroneous. In DAPA, we observed that “driving is a practical necessity in most of” Texas. 809 F.3d at 156. For that reason, we explained, it was “hardly speculative” that individuals would apply for driver’s licenses upon becoming eligible to do so. Id. at 160. This case is indistinguishable. Among other things, eligibility for a Texas driver’s license requires both residence in Texas and lawful status. And under Texas law, immigration parole under § 1182(d)(5) suffices. See Part II.C.2.b, infra pages 54–57 (explaining this). Thus, just as in DAPA, it is here “hardly speculative” that many newly paroled individuals will apply for Texas licenses. 809 F.3d at 160. Further, the district court found—with support from the record—that Texas incurs a cost for each driver’s license application it reviews. Biden I, 2021 WL 3603341, at *10 (citing a declaration of the Chief of the Texas Department of Public Safety Driver License Division, which explains, “DPS estimates that for an additional 10,000 driver['s] license customers seeking a limited term license, DPS would incur a biennial cost of approximately $2,014,870.80″). And of course, the record shows the State incurs a cost for actually granting licenses. c. Finally, the district court found that the increase in releases and paroles will increase the States’ healthcare costs. See Ibid. (citing a record deposition for the proposition that “[t]he total costs to the State will increase as the number of aliens within the state increases”).[9] That’s because both Texas and Missouri subsidize healthcare for immigrants, regardless of immigration status. See ibid. Federal law affirmatively requires the States to make some of those expenditures. See 42 C.F.R. § 440.255(c) (Emergency Medicaid). The Government appears to concede the obvious—that if the total number of in-State aliens increases, the States will spend more on healthcare. The Government’s objection, instead, boils down to repeating its claim that MPP’s termination can’t have caused either an increase in entries or an increase in parolees. Because those district court findings were not clearly erroneous, this objection goes nowhere. 2. To establish standing, the States “must show an injury that is ‘concrete, particularized, and actual or imminent; fairly traceable to the challenged action; and redressable by a favorable ruling.’” DAPA, 809 F.3d at 150 (quoting Clapper v. Amnesty Int’l USA, 568 U.S. 398, 409 (2013)). And because there was a trial, the States “needed to prove standing by a preponderance of the evidence.” Env’t Tex. Citizen Lobby, 968 F.3d at 367. Texas and Missouri each contend they have standing. But because only one of the States must have standing, we focus on Texas. See Massachusetts v. EPA, 549 U.S. 497, 518 (2007); accord NRA v. McCraw, 719 F.3d 338, 344 n.3 (5th Cir. 2013). We begin with (a) the special solicitude that Texas is owed in the standing analysis. Then we hold Texas (b) incurred an injury in fact that (c) was traceable to the Termination Decision, and that (d) can be redressed by a favorable judicial decision. Finally, we hold (e) the Government’s counterarguments are foreclosed by precedent. a. At the outset, we note that Texas is entitled to “special solicitude” in the standing analysis. Massachusetts, 549 U.S. at 520; see also DAPA, 809 F.3d at 151 (beginning with the special-solicitude question). Special solicitude has two requirements: (1) the State must have a procedural right to challenge the action in question, and (2) the challenged action must affect one of the State’s quasi-sovereign interests. Id. at 151–52 (citing Massachusetts, 549 U.S. at 516– 20). In both Massachusetts and DAPA, the first prong was satisfied because a State challenged an agency action as invalid under a statute. 549 U.S. at 516– 17 (Clean Air Act); 809 F.3d at 152–53 (APA). And in both cases, the second prong was satisfied because a State’s challenge involved an agency’s alleged failure to protect certain formerly “sovereign prerogatives [that] are now lodged in the Federal Government.” Massachusetts, 549 U.S. at 519–20; see also DAPA, 809 F.3d at 152–54. Particularly relevant here is DAPA, where we held that DAPA, by authorizing the presence of many previously unlawful aliens in the United States, affected “quasi-sovereign interests by imposing substantial pressure on them to change their laws, which provide for issuing driver’s licenses to some aliens and subsidizing those licenses.” 809 F.3d at 153 (quotation omitted). This case is no different. First, just as in the DAPA suit, Texas is asserting a procedural right under the APA to challenge an agency action. See id. at 152 (“In enacting the APA, Congress intended for those ‘suffering legal wrong because of agency action’ to have judicial recourse, and the states fall well within that definition.” (quoting 5 U.S.C. § 702)). And second, Texas asserts precisely the same driver’s-license-based injury here that it did there. See id. at 153–54 (explaining that DAPA, by greatly increasing the class of people to whom existing Texas law would entitle a subsidized driver’s license, pressured Texas to change its own law—thus affecting a quasi- sovereign interest). Thus, Texas is entitled to special solicitude in the standing inquiry. If nothing else, that means imminence and redressability are easier to establish here than usual. See Massachusetts, 549 U.S. at 517–18 (holding a State “can assert [its] right[s] without meeting all the normal standards for redressability and immediacy” (quotation omitted)).[10] b. Texas has suffered actual injury already, and it faces additional costs if the district court’s injunction ends. MPP’s termination has increased the number of immigrants paroled into Texas under 8 U.S.C. § 1182(d)(5). And as DAPA discussed at length, Texas law requires the issuance of a license to any qualified person—including aliens who “present . . . documentation issued by the appropriate United States agency that authorizes the applicant to be in the United States.” 809 F.3d at 155 (alteration in original) (quoting TEX. TRANSP. CODE § 521.142(a)); see also TEX. TRANSP. CODE § 521.181. Parole under 8 U.S.C. § 1182 satisfies that requirement. See TEX. DEP’T OF PUB. SAFETY, VERIFYING LAWFUL PRESENCE 4 (2013), https://perma.cc/Z55H-GHBH (listing an acceptable document for “parolees” as “[i]mmigration documentation with an alien number or I-94 number,” and going on to explain that “[t]his can include but is not limited to an I-94 with annotation ‘parole’ or ‘paroled pursuant to [8 U.S.C. § 1182(d)(5)]‘”); see also TEX. DEP’T OF PUB. SAFETY, U.S. CITIZENSHIP OR LAWFUL PRESENCE REQUIREMENT (2021), https://perma.cc/5AWR-HVPF (including a hyperlink to the VERIFYING LAWFUL PRESENCE document). Likewise, parole (or any other form of release into the state, as opposed to return to Mexico) satisfies Texas’s residency requirement for driver’s licenses. See TEX. TRANSP. CODE § 521.1426(a) (“The department may not issue a driver’s license or a personal identification certificate to a person who has not established a domicile in this state.”). Because driving is a “practical necessity in most of the state,” there’s “little doubt” many newly paroled aliens have applied—and without the district court’s injunction, will apply in the future—for Texas driver’s licenses. See DAPA, 809 F.3d at 156. And the district court found, without a hint of clear error, that each granted license (and each reviewed application for a license, even if not granted) costs Texas money. It follows that Texas has been actually injured—or at the least, that it faces imminent injury without the district court’s injunction. Likewise with healthcare costs. The Government says that’s not enough because Texas has not shown it has already issued any licenses to immigrants who became eligible because of MPP’s termination. Tellingly, however, it offers no hint as to how Texas could make that showing—nor why we should require it to do so. Imagine Texas had produced copies of driver’s license applications from paroled aliens. Would that have counted as evidence that Texas had, in the Government’s words, “issued a single additional driver’s license as a result” of MPP’s termination? Of course not: There would always remain some possibility that any given parolee would have been paroled even under MPP. MPP is precisely the sort of large-scale policy that’s amenable to challenge using large-scale statistics and figures, rather than highly specific individualized documents. And Texas’s standing is robustly supported by just such big-picture evidence. There is nothing “conjectural” or “hypothetical” about that. Steel Co. v. Citizens for a Better Env’t, 523 U.S. 83, 103 (1998) (quotation omitted); cf. DAPA, 809 F.3d at 161–62 (“The state must allege an injury that has already occurred or is certainly impending; it is easier to demonstrate that some DAPA beneficiaries would apply for licenses than it is to establish that a particular alien would.” (quotation omitted)). To the contrary, given both MPP’s effect of increasing the number of parolees and the fact that many of those parolees will apply for Texas licenses, it’s impossible to imagine how the Government could terminate MPP without costing Texas any money. See Clapper, 568 U.S. at 409 (“[T]hreatened injury must be certainly impending to constitute injury in fact.” (emphasis omitted)). And in all events, Massachusetts countenanced a far less obvious injury than this one. 549 U.S. at 522–23. Second, the Government resorts to Crane v. Johnson, 783 F.3d 244 (5th Cir. 2015), where this court held Mississippi lacked standing to challenge the Deferred Action for Childhood Arrivals (“DACA”) program. Id. at 252. Mississippi produced neither “evidence that any DACA eligible immigrants resided in the state,” nor “evidence of costs it would incur if some DACA-approved immigrants came to the state.” Ibid. Instead, Mississippi cited nothing more than a nine-year-old study regarding the costs of illegal immigration as a whole (not the costs imposed by DACA in particular). Id. at 249, 252. We concluded that “Mississippi’s claim of injury [was] not supported by any facts.” Id. at 252. This case is worlds apart. Texas has, of course, supported its claim of injury with facts. And that includes precisely the kind of facts Mississippi was missing: “evidence of costs it would incur” if MPP increased the number of parolees in the state. See ibid.; Biden I, 2021 WL 3603341, at *9–10 (citing record evidence of projected costs to issue additional driver’s licenses, projected costs to evaluate additional driver’s license applications, and projected healthcare costs). Third, the Government points out that there’s been a full bench trial here, unlike the preliminary-injunction posture of DAPA. That is a distinction, and it means Texas must show standing by a preponderance of the evidence rather than that it’s merely “likely” to establish standing. See Lujan v. Defs. of Wildlife, 504 U.S. 555, 562 (1992) (concluding the standing burden of proof varies with the stages of litigation); Fenves, 979 F.3d at 329 (explaining the standard at the preliminary-injunction stage). Yet the distinction changes nothing. The district court’s factual findings are not clearly erroneous. And as just explained, those findings do indeed suffice to show Texas’s actual or imminent injury by a preponderance of the evidence. Finally, the Government says Texas’s injuries are self-inflicted and therefore entirely irrelevant to the standing inquiry. See Pennsylvania v. New Jersey, 426 U.S. 660, 664 (1976) (per curiam). Our court addressed and rejected precisely this argument in DAPA. See 809 F.3d at 157–60 (citing Wyoming v. Oklahoma, 502 U.S. 437 (1992)). The Government does not acknowledge that exhaustive, precedent-based treatment of the issue, and it offers no reason at all for holding that Texas’s injury is self-inflicted in this case when it was not in DAPA. Here, as there, Texas is injured by the “Hobson’s choice of spending millions of dollars to subsidize driver’s licenses or changing its statutes.” Id. at 163. c. Texas’s injury is also traceable to DHS’s termination of MPP. The district court found that MPP’s termination has caused, and will continue to cause, an increase in immigrants paroled into Texas. Many new parolees are certain to apply for driver’s licenses—and evaluating each application will impose costs on Texas. Cf. DAPA, 809 F.3d at 160 (noting that new immigrants—in that case, DAPA recipients—”have strong incentives to obtain driver’s licenses, and it is hardly speculative that many would do so if they became eligible”). Not to mention actually granting licenses. Likewise, at least some MPP-termination-caused immigrants will certainly seek healthcare services from the State. The causal chain is easy to see. See Massachusetts, 549 U.S. at 523 (finding traceability where the EPA’s challenged action may have caused people to drive less fuel-efficient cars, which may in turn contribute to a prospective rise in sea levels, which may in turn cause the erosion of Massachusetts’s shoreline). The Government nonetheless argues that, when “a causal relation between injury and challenged action depends upon the decision of an independent third party . . . standing is not precluded, but it is ordinarily substantially more difficult to establish.” California v. Texas, 141 S. Ct. 2104, 2117 (2021) (quotation omitted). And the district court’s causal reasoning relies on mere speculation about “complex decisions made by non-citizens . . . before they risk[] life and limb to come here.” Arpaio v. Obama, 797 F.3d 11, 21 (D.C. Cir. 2015). Thus, says the Government, Texas’s injury (if any) can be traced back to immigrants’ choices, not to MPP’s termination. But the court was not speculating. It did not merely prognosticate that, sometime in the future, MPP’s termination would influence aliens’ decisions whether to immigrate illegally. Instead, the court surveyed the record and found the relevant cause-and-effect had already been taking place (even if some of its impacts on Texas were still imminent rather than actual). See Biden I, 2021 WL 3603341, at *9–10. In other words, MPP’s termination has already increased the rate of illegal entries and the number of parolees. That means the States have met their burden “to adduce facts showing that [the choices of the relevant third parties] have been or will be made in such manner as to produce causation.” Lujan, 504 U.S. at 562. Those same findings of past and present facts differentiate this case from others where the Supreme Court has refused to base standing on speculation about the future choices of third parties. See, e.g., California, 141 S. Ct. at 2118–19 (“The state plaintiffs have failed to show that the challenged minimum essential coverage provision, without any prospect of penalty, will harm them by leading more individuals to enroll in these programs.” (emphasis added)); Allen v. Wright, 468 U.S. 737, 758 (1984) (“[I]t is entirely speculative . . . whether withdrawal of a tax exemption from any particular school would lead the school to change its [racially discriminatory] policies. It is just as speculative whether any given parent of a child attending such a private school would decide to transfer the child to public school as a result of any changes in educational or financial policy made by the private school once it was threatened with loss of tax-exempt status.” (citation omitted)); Clapper, 568 U.S. at 413 (“[E]ven if respondents could show that the Government will seek the Foreign Intelligence Surveillance Court’s authorization to acquire the communications of respondents’ foreign contacts . . . respondents can only speculate as to whether that court will authorize such surveillance.” (emphasis added)). Here, unlike in those cases, MPP’s termination has already increased the rate of illegal entries into Texas. The only relevant third-party choice that remains, then, is the alien’s choice to apply for a license once in Texas.